TY - JOUR
T1 - Preliminary results of computer-aided diagnosis for magnetic resonance imaging of solid breast lesions
AU - Yu, Qiujie
AU - Huang, Kuan
AU - Zhu, Ye
AU - Chen, Xiaodan
AU - Meng, Wei
N1 - Publisher Copyright:
© 2019, Springer Science+Business Media, LLC, part of Springer Nature.
PY - 2019/9/15
Y1 - 2019/9/15
N2 - Purpose: The present study aimed to determine suitable optimal classifiers and investigate the general applicability of computer-aided diagnosis (CAD) to compare magnetic resonance (MR)-CAD with MR imaging (MRI) in distinguishing benign from malignant solid breast masses. Methods: We analyzed a total of 251 patients (mean age: 44.8 ± 12.3 years; range: 21–81 years) with 274 breast masses (154 benign masses, 120 malignant masses) using a Gaussian mixture model and a random forest machine model for segmentation and classification. Results: The diagnostic performance of MRI alone and MRI plus CAD were compared with respect to sensitivity, specificity, and area under the curve (AUC), using receiver operating characteristic curve analysis. The discriminating power to detect malignancy using MR-CAD with an AUC of 0.955 (sensitivity was 95.8% and the specificity was 92.9%) was significantly higher than that of MRI alone with an AUC of 0.785 (sensitivity was 71.7% and the specificity was 85.7%). Conclusion: CAD is feasible to differentiate breast lesions, and it can complement MRI, thereby making it easier to diagnose breast lesions and obviating the need for unnecessary biopsies.
AB - Purpose: The present study aimed to determine suitable optimal classifiers and investigate the general applicability of computer-aided diagnosis (CAD) to compare magnetic resonance (MR)-CAD with MR imaging (MRI) in distinguishing benign from malignant solid breast masses. Methods: We analyzed a total of 251 patients (mean age: 44.8 ± 12.3 years; range: 21–81 years) with 274 breast masses (154 benign masses, 120 malignant masses) using a Gaussian mixture model and a random forest machine model for segmentation and classification. Results: The diagnostic performance of MRI alone and MRI plus CAD were compared with respect to sensitivity, specificity, and area under the curve (AUC), using receiver operating characteristic curve analysis. The discriminating power to detect malignancy using MR-CAD with an AUC of 0.955 (sensitivity was 95.8% and the specificity was 92.9%) was significantly higher than that of MRI alone with an AUC of 0.785 (sensitivity was 71.7% and the specificity was 85.7%). Conclusion: CAD is feasible to differentiate breast lesions, and it can complement MRI, thereby making it easier to diagnose breast lesions and obviating the need for unnecessary biopsies.
KW - Breast lesions
KW - Computer-aided diagnosis
KW - Gaussian mixture
KW - MRI
KW - Random forest
UR - http://www.scopus.com/inward/record.url?scp=85067684724&partnerID=8YFLogxK
U2 - 10.1007/s10549-019-05297-7
DO - 10.1007/s10549-019-05297-7
M3 - Article
C2 - 31203487
AN - SCOPUS:85067684724
SN - 0167-6806
VL - 177
SP - 419
EP - 426
JO - Breast Cancer Research and Treatment
JF - Breast Cancer Research and Treatment
IS - 2
ER -